This paper focuses on estimation of the vertical velocity of the vehicle chassis and the relative velocity between chassis and wheel. These velocities are important variables in semi-active suspension control. A model-based estimator is proposed including a Kalman filter and a non-linear model of the damper. Inputs to the estimator are signals from wheel displacement sensors and from accelerometers placed at the chassis. In addition, the control signal is used as input to the estimator. The Kalman filter is analyzed in the frequency domain and compared with a conventional filter solution including derivation of the displacement signal and integration of the acceleration signal.